Back to catalog

decision-log

Log significant decisions with structured template for future review. Includes context, options, probabilities, confidence, 2nd/3rd order effects, and scheduled review dates.

Category ⚡ Productivity
Version v1.0.0
decision-makingreviewtemplatestrategy

Decision Log

Log significant decisions using a structured template so they can be revisited later.

The template captures context, options, probabilities, confidence levels, and sets

a scheduled review date — turning one-off decisions into a compounding learning loop.

When This Skill Activates

Use this skill when Joseph:

  • Makes a non-trivial decision (technology choice, strategy, tool selection, architecture)
  • Asks to "log this decision" or "document this choice"
  • Wants to revisit a past decision
  • Asks about the decision log

Decision Directory

All decisions live at /opt/data/wiki/produce/hermes/self/decisions/ (under the Self bucket). The folder contains:

  • _TEMPLATE.md — the canonical template (copy this for new decisions)
  • README.md — index of all decisions with review dates
  • YYYY-MM-DD-slug.md — individual decision files

> Standing rule: Use the full template for ALL decisions — no shortening or compression. Every entry keeps the Confidence %, Decision triggers/tripwires, and Review section. Low-stakes decisions still fill every field.

Template Structure

The template captures:

1. Header: Date, time, physical/mental state, confidence %

2. Context: Situation, why now, what prompted the decision

3. Driving Forces: Push toward change vs pull toward status quo

4. Variables at Play: Fixed factors, uncertainties, negotiables

5. Complications / Complexity: Hidden costs, second-order wrestling

6. Options Considered: Each option with pros/cons/why-rejected

7. Possible Outcomes and Probabilities: Table with best/expected/worst

8. Expected Outcome and Reasoning: Causal chain, mental model, confidence

9. 2nd and 3rd Order Effects: Non-obvious consequences

10. Final Decision: Clear statement + decision triggers/tripwires

11. Review: Scheduled date + outcome assessment (filled at review time)

Process

Logging a New Decision

1. Read /opt/data/wiki/produce/hermes/self/decisions/_TEMPLATE.md for the canonical template

2. Ask Joseph for any missing details (especially: physical/mental state, confidence %, decision triggers)

3. Write the decision file to /opt/data/wiki/produce/hermes/self/decisions/YYYY-MM-DD-slug.md

4. Add entry to /opt/data/wiki/produce/hermes/self/decisions/README.md index table

Reviewing a Past Decision

1. Read the decision file

2. Ask Joseph: "What actually happened? Was the confidence justified? What did you miss?"

3. Fill in the "Review > Outcome Assessment" section

4. Update the README.md to mark as reviewed

Key Principles

  • Confidence matters. Always capture a percentage. It's the most important field for learning calibration.
  • State matters. Physical/mental state affects decision quality. Capture it honestly.
  • Triggers matter. Every decision needs tripwires — what would cause an early revisit?
  • Be specific about probabilities. "20% chance of X" teaches more than "might happen."
  • 2nd and 3rd order effects are where the real learning lives. Don't skip them.

Pitfalls

  • Don't log trivial decisions — this is for decisions with meaningful tradeoffs
  • Don't skip the confidence % — it's the calibration mechanism
  • Don't skip the physical/mental state — it's diagnostic for pattern recognition
  • Set realistic review dates — 6 months is the default, but some decisions need 3 months or 12
  • The review section is where the value compounds — don't let decisions sit unreviewed